MiSTer Multisystem 2 Ships, £204 and No DE10-Nano

On August 6, 2025, a hobbyist computing project that had spent the better part of a decade looking like a tangle of stacked development boards finally did the most boring, most consequential thing a piece of hardware can do: it left the warehouse. Heber, the Oxfordshire engineering outfit that has commercialized the MiSTer ecosystem for the masses, posted a shipping update stating plainly that they had "started shipping pre-orders" that day. No fireworks. No keynote. Just the quiet confirmation that the MiSTer Multisystem 2 had crossed the line from promise to product.

That sentence is the entire story, and it is also the least interesting part of the story. What matters is what is inside the box, why it took this long, what it costs, and whether the central marketing claim — "low-latency, cycle-accurate" FPGA emulation in a finished consumer shell — survives contact with the people who actually understand field-programmable gate arrays. Spoiler: mostly, it does. The Machine has read the spec sheet. Let us proceed.

The Shipping Notice Nobody Doubted

Hardware launches in the retro space have a credibility problem, and it is earned. The genre is littered with crowdfunded miniconsoles that arrived two years late, FPGA boutique boards that shipped to forty people and then evaporated, and reproduction cartridges whose ship dates functioned more as zodiac signs than commitments. So when Heber said in May 2025 that the Multisystem 2 would ship in August 2025, the correct posture was polite skepticism.

What is notable, then, is how little drama there was. A June 2025 video update from The Retro Collective — the content and community arm fronted by Neil Thomas that has been welded to this project since its inception — reported that the team still planned to "start shipping early August" and that the project was, in their words, "on track." Two months later, it was. The gap between the announced date and the actual date was measured in days, not quarters. In this corner of the industry, that is close to a miracle, and it is worth saying so before we start picking the thing apart.

The reason the schedule held is structural, and it is the same reason the Multisystem 2 is interesting at all: this is not a from-scratch FPGA design chasing first silicon. It is the industrialization of an architecture — MiSTer — that has been load-bearing for thousands of users for years. Heber was not inventing the engine. They were building a better car around it.

What Actually Changed: The FPGA Comes Home

To understand the Multisystem 2, you have to understand what the original Multisystem was, and what it was bolted onto. MiSTer, the open-source project credited to Alexey Melnikov — whom Heber's own product page names directly as the originator, designer, and creator of MiSTer FPGA — was architected around a specific, off-the-shelf piece of hardware: the Terasic DE10-Nano, an Intel Cyclone V development board. The original Multisystem was, in essence, a carrier board and case that gave the DE10-Nano a friendly face: real controller ports, video output, power, and I/O that a human being could plug into without an electrical engineering degree.

It worked. It also meant that every Multisystem was a sandwich. You bought a Multisystem board, you sourced a DE10-Nano separately, you stacked them, you added I/O daughterboards for analog video, and you ended up with a working but conspicuously DIY object. The May 2025 launch video, bluntly titled "MiSTer Multisystem 2 has Arrived!", made the headline change explicit: the DE10-Nano is no longer required, because the FPGA chip is now integrated directly onto the Multisystem 2's own board.

This is the whole ballgame. By bringing the FPGA in-house, Heber's product page can describe the system as built around a new board design with direct access to the FPGA hardware — access that the original could never have, because it was always a guest on someone else's development board. That direct access is what the company says will enable "current add-ons and future expansions not possible on the original Multisystem." Concretely, the launch video quantified the upside: roughly 40 extra pins are exposed compared with the original system-and-stack arrangement. Forty pins does not sound like a revolution. In FPGA carrier terms, it is the difference between a closed appliance and an expansion platform — enough headroom to hang cartridge connectors off the machine for reading actual physical cartridges, a feature the video explicitly floated.

The second consequence of bringing the FPGA home is presentation. Heber's product page is emphatic that the Multisystem 2 ships fully assembled and ready-to-play in a custom-designed 3D printed case. Read that as a thesis statement. The original was, for all its polish, a kit-adjacent experience. The sequel is sold as a console. You take it out of the box and it works. For a project born in the soldering-iron underground, that repositioning is the actual news.

The Numbers: Dates and Pricing

Hard figures, because this article promised them. Pricing and the launch timeline were first reported via a May 2025 post quoting the official launch announcement, and later confirmed in spirit by Heber's shipping notice. There are two SKUs, split on a single axis that will be familiar to anyone who has ever argued about CRTs at 2 a.m.: whether you want analog video output or not.

The £48 analog premium is the only real purchasing decision the lineup forces, and it is a clarifying one. If your output device is a modern flat panel or a scaler, the £204 digital SKU is the rational buy. If your output device is a Sony PVM, a consumer Trinitron, or anything with a 15kHz appetite, the £252 analog SKU is not a luxury — it is the entire reason you are buying FPGA hardware instead of running software emulation on a Raspberry Pi. The whole value proposition of accuracy collapses the moment your signal chain introduces a sloppy conversion stage. Buy the right SKU for your display, not the cheap one for your wallet.

The Spec Sheet, Line by Line

Here is the hardware, assembled from Heber's product page and the launch video, with the editorializing kept to the right-hand column where it belongs.

Two entries on that sheet deserve more than a row. The first is the 24-bit DAC. Audio is the part of retro reproduction that everyone forgets until it is wrong, and a dedicated, decent-bit-depth converter on the board itself means the machine is not punting the hardest analog problem to a USB dongle. The second is the integrated PC Engine multitap with five Joypad connectors. This is gloriously specific. NEC's PC Engine — sold in North America as the TurboGrafx-16, as Wikipedia documents in exhausting detail — treated five-player multitap support as a system feature, and building that directly into the case is the kind of decision you only make if you genuinely play Bomberman with four other people. It is the opposite of a generic spec. It is a love letter.

The USB layout — five front, two rear, seven total — is the unglamorous spec that tells you Heber thought about how the thing sits on a shelf. Front ports for controllers you plug and unplug; rear ports for storage and persistent devices you set up once. That is product design, not parts-binning.

How We Got Here: DE10-Nano and the MiSTer Decade

Context, because the Multisystem 2 did not appear from nowhere. The relevant history begins with the technology itself. A field-programmable gate array is, as Wikipedia puts it, an integrated circuit whose logic can be reconfigured after manufacturing — a chip you can rewire in software to become different hardware. That property is the philosophical heart of the entire enterprise. Software emulation imitates a Super Nintendo by running instructions that describe its behavior. An FPGA becomes a circuit that behaves like a Super Nintendo's silicon, gate for gate, clock for clock. The difference is not pedantic; it is the difference that the whole £204-to-£252 spend is predicated on.

MiSTer took this idea and democratized it. By standardizing on the Terasic DE10-Nano — an affordable, widely available Cyclone V board — Alexey Melnikov and the contributors around the project gave the community a fixed target. Write a core once, and it runs on every MiSTer setup on Earth, because every MiSTer setup is the same board. That standardization is why the core library exploded: arcade boards, home computers, consoles from the 8-bit era through the 16-bit wars, each one a downloadable circuit. The trade-off was the one we have already named — the DE10-Nano was a development board, not a consumer product, and building a livable system around it required carrier boards, I/O daughterboards, cases, and patience. The original Multisystem was Heber's answer to that friction. The Multisystem 2 is the answer to the answer.

It is worth situating this against the broader history of console emulation, which for thirty years has been a software story — SNES emulators on 90s PCs, MAME's heroic preservation campaign, the handheld emulation boom. FPGA reproduction is the counter-movement: slower, more expensive, more pedantic, and aimed at the specific cohort that cares about input latency and timing accuracy enough to pay for silicon-level fidelity. The Multisystem 2 is the most consumer-friendly artifact that movement has produced to date. That is its place in the lineage: not the first FPGA console, but the one that most convincingly looks and behaves like an appliance while still being one.

The Accuracy Claim, Interrogated

Heber's product page leans on a two-word phrase that does enormous load-bearing work: "low-latency, cycle-accurate." The page further frames the machine as delivering "the most accurate gaming experience compared with playing on original hardware." The launch video calls the system "100% compatible" with the original Multisystem concept. These are the claims that justify the price. So let us be precise about what they do and do not mean.

Cycle-accurate is a real and meaningful term of art. It means the reproduction advances its internal state on the same clock boundaries as the original hardware, so that timing-dependent behavior — raster effects, audio mixing, the precise frame on which a sprite appears — matches the original rather than approximating it. Software emulators can achieve this too (bsnes/higan famously chased it), but at a steep CPU cost and, critically, with display latency introduced by the operating system and graphics stack underneath. The FPGA approach sidesteps that stack. There is no OS scheduler between the core and the output. That is where the low-latency half of the phrase earns its keep: a properly implemented FPGA core can deliver input-to-photon latency close to the original console's, which software running on a general-purpose OS structurally cannot guarantee.

The honest caveat — and The Machine deals in honest caveats — is that "cycle-accurate" is a property of the core, not the box. The Multisystem 2 runs MiSTer cores, and the quality of those cores varies by system and by maturity. The hardware gives you the best possible substrate for accurate reproduction; it cannot retroactively make an immature core perfect. So the correct reading of Heber's claim is: the machine removes every accuracy bottleneck that the platform can remove, and what remains is the well-understood, slowly-improving quality of individual community cores. That is a strong position. It is not magic, and the company's phrasing — compared with playing on original hardware — is a comparison, not a claim of identity. Read carefully, it is defensible.

For the comparison-shopper, it is instructive to set this against the most famous FPGA consumer product, the Analogue Pocket. The Verge's review of the Analogue Pocket captured the appeal of FPGA hardware for a mainstream audience — fidelity as a feature you can feel rather than measure on a spec sheet. The difference is one of philosophy: Analogue ships closed, polished, vertically integrated products with curated FPGA cores. MiSTer, and by extension the Multisystem 2, ships an open platform with a community core library. You are buying into an ecosystem, not a sealed appliance, even though the Multisystem 2 now looks like a sealed appliance.

The Competition: Analogue, Software, and the Old Stack

There are three things a prospective Multisystem 2 buyer is realistically choosing between, and it is worth naming each.

Against the original Multisystem-plus-DE10-Nano stack. This is the easiest comparison, and the Multisystem 2 wins it on every axis except possibly total cost if you already own a DE10-Nano. The integrated FPGA, the ~40 extra pins, the assembled case, the 24-bit DAC, and the built-in multitap collectively render the old sandwich obsolete for new buyers. The only reason to build the old stack now is if you specifically need a DE10-Nano for non-MiSTer FPGA work and want to amortize it.

Against Analogue's product line. Analogue (the Pocket, and its console-form devices) is the polished, retail-channel, fixed-function competitor. The trade is clear: Analogue gives you a more designed object and a curated experience; the Multisystem 2 gives you the breadth of the open MiSTer core library and a hardware expansion path — those 40 pins, the mooted cartridge connectors — that a closed product structurally will not offer. Mainstream coverage of FPGA gaming, including the Analogue write-ups across outlets like Engadget, has tended to frame these devices as enthusiast objects, and that framing applies to both camps. The Multisystem 2 is the more open, more expandable, more hackable option for someone who already speaks MiSTer.

Against software emulation. This is the comparison that decides whether you buy any FPGA hardware at all. A modern handheld or a Raspberry Pi running RetroArch will play the same games for a fraction of the price. What it will not reliably deliver is the latency floor and the clean analog signal path that justify the Multisystem 2's existence. If you cannot perceive or do not care about input latency, and you output to a flat panel, software emulation is the rational choice and FPGA hardware is an indulgence. If you play on a CRT, chase frame-perfect inputs, or simply value silicon-level reproduction as a matter of principle, the £204 floor is the cost of that principle.

Multisystem 2 (digital SKU)      = £204  | FPGA integrated, assembled, 7x USB, 24-bit DAC
Multisystem 2 (analog SKU)       = £252  | adds analog video output for CRT chains
   analog premium                = £48

Original Multisystem stack       = board + DE10-Nano + I/O board + case + assembly time
   (multi-part, DIY-leaning, no integrated FPGA, fewer expansion pins)

Decision logic:
   flat panel only            -> £204 digital SKU
   CRT / 15kHz signal chain   -> £252 analog SKU
   indifferent to latency     -> software emulation (cheaper, OS-bound latency)

The Pocket Question: A Handheld MiSTer

The most forward-looking item in the 2025 record is not the Multisystem 2 at all — it is what The Retro Collective said about what comes next. In a 2025 update, the team stated flatly: "we're working on a handheld MiSTer," and named the ambition: a "finished Multisystem 2 Pocket THIS YEAR."

Take that seriously, but not literally on the date. A portable MiSTer is a genuinely hard engineering problem — FPGAs are power-hungry relative to the SoCs in commodity handhelds, and battery life, thermals, and a built-in display are exactly the constraints that have historically kept FPGA reproduction tethered to wall power. The fact that integrating the FPGA onto the Multisystem 2's own board was a prerequisite is telling: you cannot make a pocketable MiSTer while the architecture still demands a stacked DE10-Nano. The desktop sequel is, in this reading, the foundation that makes a handheld conceivable. The "this year" framing should be read as an aspiration from a team that just hit an August ship date, not as a binding commitment. But the direction is unambiguous, and it reframes the Multisystem 2 from an endpoint into a platform generation.

Five Predictions for the Next 6-12 Months

The Machine does not forecast lottery numbers. It extrapolates from the record. Here is what the next two to four quarters most plausibly hold, given what shipped on August 6 and what the team has said out loud.

1. Cartridge-reader add-ons move from "mooted" to "real." The launch video explicitly tied the ~40 extra pins to cartridge connectors for reading physical media. Expect at least one official or community cartridge-adapter accessory to appear within 6-12 months, because the entire point of exposing those pins was to enable exactly this, and the demand for dumping and playing original carts is bottomless.
2. The Multisystem 2 Pocket gets shown, but slips its "this year" target. A handheld MiSTer is hard for reasons of power and thermals. Expect public prototypes, a teardown or two, and revised timelines — a credible reveal, not a 2025 retail handheld. If it ships on time, treat that as the second miracle in two years.
3. Supply, not demand, governs the next two quarters. Pre-orders that opened in May and shipped in August imply a backlog. Expect the story through early 2026 to be about fulfilling that queue and stabilizing the assembled-console supply chain, not about a price cut. Assembled hardware with a 3D-printed case does not scale instantly.
4. Core development accelerates around the new I/O. Direct FPGA access and 40 new pins are an invitation. Expect community cores and forks that specifically exploit Multisystem 2-only hardware features — the multitap, the DAC, the expansion header — which will, mildly and predictably, create a soft tier of "works best on Multisystem 2" experiences.
5. Mainstream coverage finally arrives. As of the 2025 source record, the heavyweight outlets had not covered this in depth. A shipping, assembled, photogenic FPGA console with a multitap and a handheld on the roadmap is precisely the kind of object that FPGA-curious general-tech press eventually writes up. Expect a second wave of secondary coverage as units land in reviewers' hands.

The Verdict: A Kit Becomes a Console

Strip away the spec rows and the price deltas and the prediction list, and the MiSTer Multisystem 2 represents a single, clean transition: a project that was always about capability finally became a project about product. The original Multisystem made MiSTer livable. The Multisystem 2 makes it buyable — fully assembled, FPGA integrated, DE10-Nano retired, plugged in and running out of a 3D-printed box that you did not have to solder.

The accuracy claims hold up under scrutiny, with the honest asterisk that "cycle-accurate" describes the platform's ceiling and individual cores describe the floor. The pricing is rational and the SKU split is the right split. The hardware decisions — seven USB ports laid out like someone uses them, a 24-bit DAC, a PC Engine multitap built into the chassis — read as the work of people who actually play the things they build. And the 40 extra pins quietly convert a finished console into an expandable one, with cartridge reading and a handheld successor both visible on the horizon.

It shipped on August 6, 2025, days off a date announced in May. In a field where "shipping" is usually a verb in the future tense, that is the highest praise The Machine offers: it said when, and then it did. Buy the analog SKU if you own a CRT. Buy the digital SKU if you do not. And keep an eye on the Pocket — because the most interesting thing about the Multisystem 2 may turn out to be that it was the foundation, not the building.